Article

Mechanistic Basis of Organization of the Harmonin/ USH1C-Mediated Brush Border Microvilli Tip-Link Complex

Graphical Abstract Authors Jianchao Li, Yunyun He, Qing Lu, Mingjie Zhang

Correspondence [email protected] (Q.L.), [email protected] (M.Z.)

In Brief Li et al. present a systematic biochemical and structural characterization of the brush border inter-microvillar tip-link complex assembly, revealing its striking similarity with the inner ear hair cell inter- stereocilia complex organizations, despite their minimal overlap in proteins. The results provide mechanistic insight into the function of brush border microvilli under both normal physiological and disease conditions.

Highlights Accession Numbers d The brush border inter-microvilli link complex assembly is 5F3X characterized 5F3Y d The Harmonin/ANKS4B/Myo7B complex links cadherins with actin bundles d Harmonin acts as the key scaffold of the inter-microvilli tip- link complex d Despite minimal overlap, microvilli and stereocilia tip-link complexes are similar

Li et al., 2016, Developmental Cell 36, 179–189 January 25, 2016 ª2016 Elsevier Inc. http://dx.doi.org/10.1016/j.devcel.2015.12.020 Developmental Cell Article

Mechanistic Basis of Organization of the Harmonin/USH1C-Mediated Brush Border Microvilli Tip-Link Complex

Jianchao Li,1,3 Yunyun He,1,3 Qing Lu,1,2,* and Mingjie Zhang1,2,* 1Division of Life Science, State Key Laboratory of Molecular Neuroscience, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong, China 2Center of Systems Biology and Human Health, School of Science and Institute for Advanced Study, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong, China 3Co-first author *Correspondence: [email protected] (Q.L.), [email protected] (M.Z.) http://dx.doi.org/10.1016/j.devcel.2015.12.020

SUMMARY molecules that play critical roles in governing the development and maintenance of stereocilia have been identified and exten- Brush border microvilli are actin-based protrusions sively studied (El-Amraoui and Petit, 2005; Pan and Zhang, lining the apical surface of epithelial cells in intestines 2012; Schwander et al., 2010). Mutations of genes encoding and proximal tubules of kidneys. While brush border many of these molecules result in syndromic or non-syndromic microvilli resemble the relatively well-characterized deafness (Richardson et al., 2011). Among syndromic deaf- stereocilia of hair cells, the mechanistic basis of ness, Usher syndrome I is the most severe form of heritable tip-link complex organization in microvilli is poorly hearing and visual impairment disorders caused by mutations of one of five USH1 genes (Mathur and Yang, 2015; Richardson understood. Here, we have biochemically and struc- et al., 2011). Microvilli are found at the apical surface of epithe- turally characterized the following pairs of interac- lial cells such as small intestines and kidney, and form densely tions: protocadherin 24 and Harmonin (also known packed bundle-like structures called brush borders (Louvard as USH1C or AIE-75), Harmonin and myosin VIIb et al., 1992; Mooseker, 1985). Recent studies have shown (MYO7B), Harmonin and ANKS4B, and ANKS4B that many of the protein components that regulate stereocilia and MYO7B. We show that Harmonin, ANKS4B, development and maintenance have their corresponding coun- and MYO7B form a stable ternary complex for terparts in microvilli (Crawley et al., 2014b; McConnell et al., anchoring microvilli tip-link cadherins. Despite hav- 2011), suggesting that the underlying molecular mechanisms ing only Harmonin in common, the microvilli and for the development of stereocilia and microvilli might be similar the stereocilia tip-link complexes are formed via (Crawley et al., 2014a). strikingly similar interaction modes. These results In stereocilia, two Usher syndrome gene encoded cell adhe- sion proteins, protocadherin 15 (PCDH15, encoded by USH1F/ not only provide insight into the mechanistic bases PCDH15) and cadherin 23 (CDH23, encoded by USH1D/ of brush border microvilli formation and maintenance CDH23), form an inter-stereocilia linker called tip-link (Kaz- but may also be valuable for understanding some gut mierczak et al., 2007; Sotomayor et al., 2012). Microvilli also and/or kidney diseases caused by perturbations of contain inter-microvillar links located near the apex of each brush border microvilli structures. microvillus, which are formed by two cadherin family proteins known as cadherin-related family member 2 (CDHR2, also known as protocadherin-24, encoded by CDHR2) and cad- INTRODUCTION herin-related family member 5 (CDHR5, also known as mucin- like protocadherin, encoded by CDHR5)(Crawley et al., Cell surface protrusions are important to cells in many aspects 2014b). CDH23 interacts with a scaffold protein called Harmo- such as harvesting nutrients, sensing and responding to envi- nin through its cytoplasmic tail (Pan et al., 2009; Siemens et al., ronmental stimuli, and navigating movements (Nambiar et al., 2002; Wu et al., 2012). Harmonin, encoded by USH1C,isa 2010; Ridley, 2011). Based on their supporting core cytoskel- multi-PDZ domain-containing protein and expresses in three etal structures, cellular protrusions can be divided into two major spliced isoforms known as Harmonin a, b, and c (Verpy groups, microtubule-based protrusions known as cilia and et al., 2000). They all contain an N-terminal helical domain actin filament-based protrusions such as microvilli, stereocilia, (NTD), followed by two PDZ domains and a predicted coiled- and filopodia. Morphologically, microvilli and stereocilia share coil (CC) domain. The isoforms a and b contain an additional some common features. They are both assembled into PDZ domain in their C termini. Harmonin b contains another bundle-like structures (Barr-Gillespie, 2015). Stereocilia are or- predicted CC domain (CC2) and an unstructured proline-, ganelles located at the apical side of hair cells in cochlea, and serine-, and threonine-rich sequence which was reported to are important for hearing and balance (Tilney et al., 1992). Many bind to F-actin (Boeda et al., 2002). Harmonin is also known

Developmental Cell 36, 179–189, January 25, 2016 ª2016 Elsevier Inc. 179 Figure 1. Biochemical Characterization of the CDHR2 and Harmonin Interaction (A) Domain organizations of CDHR2 and Harmonin a. (B) ITC result showing that CDHR2PBM binds to

Harmonin PDZ2CC with Kd of 1.35 mM. (C) FPLC-based assay also showing that CDHR2PBM interacts with Harmonin PDZ2CC in solution. (D) Summary of the binding affinities between different fragments of CDHR2 and Harmonin derived from ITC-based assays, showing that PBM and PDZ2CC were the minimal binding regions of the complex.

et al., 2004). Thus, it is possible that ANKS4B may act as an adaptor protein between Harmonin and MYO7B in brush as autoimmune enteropathy-related 75-kDa antigen (AIE-75), border, similarly as USH1G does for linking Harmonin and and is highly expressed in the intestine and kidney (Kobayashi MYO7A in stereocilia. et al., 1998, 1999). Autoimmune enteropathy is a rare immune Here, using a combination of structural and biochemical ap- system disorder characterized by severe diarrhea and weight proaches, we show that one of the components of the inter- loss from malnutrition absorption (Kobayashi et al., 1999). Inter- microvillar link, CDHR2, uses its cytoplasmic tail PBM to bind estingly, it has been reported that the Usher syndrome patients to the PDZ2-CC region of Harmonin. We demonstrate that with mutations in Harmonin also suffered from severe gas- ANKS4B acts as an adaptor linking Harmonin and MYO7B. trointestinal symptoms and renal tubular dysfunction (Bitner- The C-terminal SAM domain and PBM of ANKS4B together Glindzicz et al., 2000), indicating that Harmonin may also play interact with the Harmonin NTD-PDZ1 supramodule, and the important roles in formation and function of intestine and kid- structure of the complex solved here reveals the atomic details ney brush border microvilli. Mirroring CDH23, CDHR2 has of the interaction. We further show biochemically and structurally been reported to use its C-terminal PDZ binding motif (PBM) that the central region of ANKS4B binds to the N-terminal to interact with Harmonin PDZ in brush border microvilli (Craw- MyTH4-FERM-SH3 supramodule of MYO7B with a mechanism ley et al., 2014b), although the detailed interaction mode is not highly analogous to the interaction between USH1G and clear. In stereocilia, Harmonin forms a very stable tertiary com- MYO7A. Finally, we demonstrate that an extended PDZ3 of plex with MYO7A (also known as unconventional myosin VIIa, Harmonin binds to the C-terminal MyTH4-FERM tandem of encoded by USH1B) through the adaptor protein USH1G MYO7B, revealing an unexpected binding mode both for PDZ (also known as sans, encoded by USH1G)(Grati and Kachar, domains and for MyTH4-FERM tandems. Our study represents 2011; Pan and Zhang, 2012). MYO7A is scarce in microvilli. a comprehensive mechanistic characterization of the inter- Instead its close homolog, MYO7B, is abundantly expressed microvillar tip-link complex assembly in the cytoplasmic face. in brush border microvilli (Chen et al., 2001; McConnell et al., The results presented in this study will be valuable for under- 2011). Again, mirroring stereocilia, Harmonin has been found standing the development and maintenance of brush border to associate with MYO7B (Crawley et al., 2014b). Curiously, microvilli and for understanding gastrointestinal and/or kidney microvilli are not known to express the adaptor protein diseases resulting from brush border microvilli defects due to USH1G, and thus how the membrane-anchored CDHR2/Har- malfunctioning of the proteins characterized herein. monin complex is linked with the actin filament-attached MYO7B is not clear. RESULTS We have noted with interest that several epithelial tissues contain a protein highly similar to USH1G, called ankyrin repeat CDHR2 Interacts with Harmonin through Its and SAM domain-containing protein 4B (ANKS4B, encoded by C-Terminal PBM ANKS4B, also known as harmonin-interacting, ankyrin repeat- A recent breakthrough study has shown that CDHR2 and containing protein, harp). ANKS4B was originally identified in CDHR5 interact with each other to form the inter-microvillar kidney (Johnston et al., 2004) and shares high similarity with link located at the apex of intestine brush border microvilli, and USH1G, both in their sequences (40% sequence identity) and the cytoplasmic tails of these two cadherins associate with Har- domain organizations. The Human Protein Atlas (http://www. monin (Crawley et al., 2014b). We first sought to characterize the proteinatlas.org/) showed that ANKS4B has the highest expres- detailed interactions governing the connection between Harmo- sion in intestine. The prominent expression of ANKS4B in intes- nin and the cadherins. The cytoplasmic tail of CDHR2 contains a tine has also been confirmed by a recent proteomics study conserved PBM (Figure 1A), and the cytoplasmic tail of CDHR5 is (Uhlen et al., 2015). Similar to USH1G, ANKS4B contains a not conserved. Thus, we reasoned that the cytoplasmic domain type I PBM in its C terminus, and was reported to interact (CD) of CDHR2 may directly bind to Harmonin, and CDHR5 may with Harmonin PDZ1 in a PBM-dependent manner (Johnston not interact with Harmonin. Indeed, we found, using purified

180 Developmental Cell 36, 179–189, January 25, 2016 ª2016 Elsevier Inc. Table 1. Details of the Proteins and Their Boundaries Used in the Crawley et al. (2014b) showing that PDZ2 alone could not Study be pulled down by CDHR2CD. Finally, a 17-residue synthetic Amino Acid peptide corresponding to the CDHR2PBM but missing the last Protein Name Residue Boundary residue (i.e. residues 1,291–1,307 of CDHR2, referred to as CDHR2 Mus musculus (Uniprot: E9Q7P9) CDHR2PBM*) had no detectable binding to Harmonin PDZ2CC (Figure 1D), indicating that CDHR2PBM is indispensable for CDHR2CD 1,176–1,308 the CDHR2/Harmonin interaction. In conclusion, similar to the CDHR2PBM 1,291–1,308 interaction between CDH23 PBM and Harmonin PDZ2 in stereo- CDHR2PBM* 1,291–1,307 cilia, CDHR2 PBM also binds to Harmonin PDZ2, albeit that Harmonin a Homo sapiens (Uniprot: Q9Y6N9) the interaction between CDHR2 and Harmonin also requires NPDZ1 1–194 the CC region following Harmonin PDZ2. At this stage, we do PDZ2 208–299 not understand how the CC region enhances Harmonin PDZ2CC 193–383 PDZ2’s binding to CDHR2CD as we have not been able to obtain diffracting-quality crystals of the Harmonin PDZ2CC/ CC 300–383 CDHR2PBM complex. The CC region may directly participate NC-extended-PDZ3 428–552 in the PDZ2CC/CDHR2PBM interaction. Alternatively, the CC re- PDZ3 N extension 428–449 gion may simply facilitate folding and stability of Harmonin PDZ2, PDZ3 C extension 537–552 forming an extended PDZ domain (Wang et al., 2010; Ye and N-extended PDZ3 428–542 Zhang, 2013). ANKS4B Mus musculus (Uniprot: Q8K3X6) SAM-PBM 345–423 ANKS4B Binds to Harmonin in a Mode Highly Similar to CEN 253–352 that of USH1G CEN1 253–330 ANKS4B, similar to USH1G, consists of four N-terminal ANK re- peats, a central region, and a sterile a motif (SAM) followed by CEN2 331–352 a C-terminal type I PBM (Figure 2A). In a previous study, the MYO7B crystal structure of Harmonin NPDZ1/USH1G SAM-PBM com- NMFS Mus musculus 962–1,578 plex showed that Harmonin NTD, PDZ1, and a short C-terminal (Uniprot: Q99MZ6) extension form a supramodule to interact with USH1G SAM- CMF Homo sapiens 1,582–2,116 PBM with very high affinity (Yan et al., 2010). Taking into (Uniprot: Q6PIF6) account the high similarity between ANKS4B and USH1G (Fig- MYO7A Mus musculus (Uniprot: P97479-2) ure 2D), we wondered whether ANKS4B SAM-PBM might also NMFS 965–1,649 interact with Harmonin NPDZ1 using the same binding mode. CMF 1,650–2,215 Since NPDZ1 tends to form aggregate in low-salt concentration USH1G Homo sapiens (Uniprot: Q495M9) buffer (e.g. 100 mM NaCl) and the protein adopts as a stable CEN 295–390 monomer in high-salt concentration buffers (Yan et al., 2010), all the experiments involving NPDZ1 described in this work CDHR5 Mus musculus (Uniprot: Q8VHF2) were performed in buffer containing 500 mM NaCl. As such, CD 663–831 the corresponding bindings are likely to be even stronger if *indicates mutation of PBM by deleting the last residue. measured under physiological salt concentrations. ITC-based assay indeed detected a very strong interaction between Har- recombinant proteins, that the entire CDHR2CD binds to the full- monin NPDZ1 and ANKS4B SAM-PBM with Kd of 3nM(Fig- length Harmonin with a dissociation constant (Kd)of5.7 mM ure 2B), which is comparable with the interaction between (Figures 1D and S1A; the construct details of all recombinant Harmonin NPDZ1 and USH1G SAM-PBM (Kd 1.3 nM) (Yan proteins used in this study are summarized in Table 1 for both et al., 2010). FPLC coupled with static light scattering (FPLC- clarity and easy comparison), a value typical for a canonical SLS) assay demonstrated that Harmonin NPDZ1 and USH1G PDZ/PBM interaction (Ye and Zhang, 2013). In contrast, no inter- SAM-PBM form a stable complex with a 1:1 stoichiometry action could be detected between CDHR5 CD and Harmonin (Figure 2C). (Figure S1B). Further detailed mapping experiments revealed To elucidate the atomic basis of the Harmonin/ANKS4B inter- that CDHR2 PBM binds to the PDZ2 and CC regions (PDZ2CC) action, we solved the crystal structure of the Harmonin NPDZ1/ of Harmonin (Figures 1B–1D). Analytical gel filtration chromatog- ANKS4B SAM-PBM complex at 2.65 A˚ resolution (Table S1). The raphy (fast protein liquid chromatography [FPLC]) assay also PBM of ANKS4B inserts into the groove formed by aB and bBof showed that CDHR2 PBM and Harmonin PDZ2CC can interact Harmonin PDZ1, and the ANKS4B SAM domain packs with the with each other (Figure 1B). The isothermal titration calorimetry exposed side of aB of Harmonin PDZ1 (Figure 2E). In addition (ITC)-based assay showed that CDHR2 PBM binds to Harmonin to the expected interactions between the ‘‘0’’ and ‘‘À2’’ positions

PDZ2CC with a Kd of 1.4 mM(Figure 1C). Removal of CC from of ANKS4B PBM and Harmonin PDZ1, the invariable Asp420PBM PDZ2CC weakens the binding to a Kd of 38 mM, and the CC at the ‘‘À3’’ position forms a salt bridge with Arg103NPDZ1 domain alone has no detectable binding to CDHR2CD (Fig- (Figure 2F). The interaction between ANKS4B SAM domain ure 1D). The weak interaction between Harmonin PDZ2 and and Harmonin PDZ1 is mediated via two pairs of salt bri-

CDHR2CD is consistent with the result in the previous study by dges (Lys407SAM-Glu148PDZ1, Arg408SAM-Glu149PDZ1), several

Developmental Cell 36, 179–189, January 25, 2016 ª2016 Elsevier Inc. 181 Figure 2. Biochemical and Structural Characterization of the ANKS4B and Harmonin Interaction (A) Domain organizations of Harmonin, ANKS4B, and USH1G.

(B) ITC result showing that ANKS4B SAM-PBM binds to Harmonin NPDZ1 with a very strong affinity (Kd 3 nM). (C) FPLC-SLS analysis showing that ANKS4B SAM-PBM and Harmonin NPDZ1 formed a stable, 1:1 stoichiometric complex in solution. (D) Sequence alignment of ANKS4B SAM-PBM and USH1G SAM-PBM from different vertebrate species. The symbols above the sequences are defined as follows: an asterisk (*) indicates positions with a fully conserved residue; a colon (:) indicates conservation between groups of strongly similar properties; a period (.) indicates conservation between groups of weakly similar properties. (E) Overall structure of the ANKS4B SAM-PBM/Harmonin NPDZ1 complex. (F) Combined stick and ribbon models showing the interaction details between ANKS4B SAM-PBM and Harmonin NPDZ1.

hydrogen bonds (side chain of Gln361SAM and side chain of An earlier study reported that Harmonin NPDZ1 (denoted Asn152PDZ1, side chain of His362SAM and backbone carbonyl ‘‘PDZ1’’ by the authors), but not PDZ2 and PDZ3, can bind to of Thr156PDZ1, side chain of Arg408SAM and backbone carbonyl CDHR2CD (Crawley et al., 2014b). We tried to verify these results of Ser144PDZ1), and hydrophobic interactions (Leu412SAM- using purified proteins. ITC-based assay of the interaction be- Leu106PDZ1, Leu418SAM-Phe108PDZ1)(Figure 2F). The combined tween CDHR2CD and Harmonin NPDZ1 in the high-salt buffer PDZ/PBM and PDZ/SAM interactions presumably lead to the detected a Kd of 22 mM(Figure S3A), suggesting NPDZ1 can formation of the highly stable Harmonin/ANKS4B complex. The indeed bind, albeit weakly, to CDHR2CD (about 20 times weaker residues from ANKS4B SAM-PBM that are involved in the inter- than the CDHR2CD/PDZ2CC interaction assayed in buffer con- action with Harmonin NPDZ1 can also be found in the corre- taining 100 mM NaCl, Figure 1B). However, it is possible that the sponding positions in USH1G SAM-PBM (Yan et al., 2010) high concentration of salt in the assay buffer might have weak- (Figure 2D). Not surprisingly, the overall structure of the Harmo- ened the binding. To verify this possibility, we quantified the nin NPDZ1/ANKS4B SAM-PBM complex is essentially the same binding affinities of CDHR2CD to PDZ2CC in the assay buffer as that of the Harmonin NPDZ1/USH1G SAM-PBM complex containing 500 mM NaCl, and the resulting Kd was 20 mM(Fig- (root-mean-square deviation of 0.71 A˚ , Figure S2). ures S3B and S3C). Thus, NPDZ1 and PDZCC of Harmonin can

182 Developmental Cell 36, 179–189, January 25, 2016 ª2016 Elsevier Inc. Figure 3. The Central Region of ANKS4B Binds to the N-Terminal MyTH4-FERM Tandem of MYO7B (A) Domain organizations of ANKS4B, MYO7B, and their respective stereocilia counterparts USH1G and MYO7A. (B) FPLC-based assay showing that ANKS4B CEN and MYO7B NMFS can interact with each other in vitro. (C) Sequence alignment of the CEN regions of ANKS4B and USH1G from different species, showing that the boxed CEN1 and CEN2 regions are both highly conserved, whereas their flanking sequences are variable. The symbols above the sequences are the same as those defined in Figure 2D. (D) ITC-based assay reveals that the full-length ANKS4B and MYO7B NMFS interact with each other with moderate affinity. (E) ITC result shows that ANKS4B CEN had a similar binding affinity toward MYO7B NMFS compared with the full-length ANKS4B. (F) Summary of ITC-derived binding affinities of different fragments of ANKS4B, MYO7B, USH1G, and MYO7A, showing that the ANKS4B/MYO7B and the USH1G/MYO7A interactions are each very specific. bind to CDHR2CD with comparable affinities. However, the MYO7B N-Terminal MyTH4-FERM-SH3 Specifically NPDZ1/SAM-PBM structure shows that the PBM-binding Interacts with the ANKS4B CEN Region groove of Harmonin NPDZ1 is occupied by ANKS4B SAM- In stereocilia, the USH1G CEN region interacts with the MYO7A PBM, and the interaction between Harmonin NPDZ1/ANKS4B N-terminal MyTH4-FERM domain (Wu et al., 2011), and the inter- SAM-PBM is nearly 104-fold stronger than the Harmonin action serves to link the CDH23/Harmonin/USH1G complex with NPDZ1/CDHR2CD interaction. Therefore, it is predicted that the actin filament bundles via MYO7A’s motor head. We first CDHR2 will not bind to Harmonin NPDZ1 in the presence of compared the amino acid sequences of the central region of ANKS4B. Indeed, no binding could be detected when ANKS4B with those of USH1G (Figures 3A and 3C). Strikingly, CDHR2PBM was used to titrate the NPDZ1/SAM-PBM complex the previously identified CEN1 and CEN2 regions of USH1G, (Figure S3D). In contrast, the binding of CDHR2CD to the full- which are directly responsible for binding to the MYO7A length Harmonin complex is not affected by the presence of MyTH4-FERM tandem (Wu et al., 2011), are both highly ANKS4B SAM-PBM (Figure S3E). Therefore, we believe that conserved in ANKS4B (Figure 3C). MYO7B is the paralog of PDZ2CC of Harmonin is the bona fide binding site for CDHR2CD MYO7A (54% sequence identity), and is expressed primarily in in brush border microvilli. kidney and intestine (Chen et al., 2001). Similar to MYO7A,

Developmental Cell 36, 179–189, January 25, 2016 ª2016 Elsevier Inc. 183 Figure 4. Structure of the ANKS4B CEN/ MYO7B NMFS Complex (A) Electron density of ANKS4B CEN1 in the

ANKS4B CEN/MYO7B NMFS complex. The Fo-Fc map was calculated by omitting CEN1 from the final model and contoured at 3s. (B) Detailed interactions between ANKS4B CEN1 (with the stick model in blue) and MYO7B NMFS. (C and D) Comparison of the ANKS4B CEN/ MYO7B NMFS and the USH1G CEN/MYO7A NMFS structures, showing that ANKS4B and USH1G adopt similar binding modes toward the respective MYO7 NMFS.

To further understand the interaction between ANKS4B CEN and MYO7B NMFS, we determined the crystal struc- ture of the MYO7B NMFS/ANKS4B CEN complex at 3.4 A˚ resolution (Table S1). The overall structure of MYO7B NMFS adopts a Y-shaped conformation, in which the MyTH4 domain physically inter- MYO7B can be divided into three parts: the motor head, the neck acts with FERM-F1 lobe, forming an intimately packed MyTH4- region, and the tail cargo-binding domain. Curiously, MYO7B FERM tandem as observed in previously characterized myosin lacks the so-called single a helix that is found in MYO7A MyTH4-FERM tandems (Figure 4C) (Hirano et al., 2011; Wei following its five IQ motifs. The tail region of MYO7B also con- et al., 2011; Wu et al., 2011). The SH3 domain also contacts sists of two MyTH4-FERM tandems separated by an SH3 the FERM-F3 lobe, although the relative position is different domain (Figure 3A). Sequence alignment analysis showed that from that in MYO7A (the SH3 domain of MYO7B rotates by N-terminal MyTH4-FERM-SH3 (NMFS) of MYO7A and MYO7B 17 clockwise when the F3 lobes are superimposed, Figure 4C). are highly similar, and most of the residues that form the hydro- Near the center of the ‘‘cloverleaf’’ of the FERM domain, the phobic CEN1 binding groove in MYO7A NMFS can also be found strong positive electron density allows us to build a b-hairpin in MYO7B NMFS (Figure S4). Therefore, we hypothesized that structure corresponding to ANKS4B CEN1 (Figure 4A). The bind- MYO7B NMFS might also be able to bind to the ANKS4B CEN ing between ANKS4B CEN1 and MYO7B NMFS is dominated by region. hydrophobic interactions. Side chains of Ile264ANKS4B and To test this hypothesis, we used ITC to measure the binding of Phe275ANKS4B from ANKS4B CEN1 make contacts with side the full-length ANKS4B to NMFS or C-terminal MyTH4-FERM chains of Leu1385MYO7B, Leu1390MYO7B, and Leu1418MYO7B (CMF) of MYO7B. We found that the NMFS binds to the full- from NMFS F3 lobe (Figure 4B). In addition, two bulky residues, length ANKS4B with a Kd of 1.1 mM(Figure 3D), whereas no Phe1227MYO7B and Trp1228MYO7B from NMFS F1 lobe, interact binding was detected between CMF and ANKS4B (Figure 3F). with Leu270ANKS4B, Ile273ANKS4B, and Val274ANKS4B (Figure 4B). Next, we designed a fragment of ANKS4B CEN (amino acids The conformations and positions of ANKS4B CEN1 and USH1G 253–352, Figure 3C) with its boundary matching that of the CEN1 are very similar in their respective NMFS-bound state (Fig- USH1G CEN region (Wu et al., 2011). ITC assay showed that ures 4C and 4D). In the structure of MYO7A NMFS/USH1G CEN

ANKS4B CEN binds to MYO7B NMFS with a Kd of 1.7 mM(Fig- complex, the structure of CEN2 cannot be reliably determined by ure 3E), indicating that the CEN region is sufficient for the X-ray crystallography. However, sparse electron densities can ANKS4B/MYO7B interaction. In the FPLC study, the 1:1 mixture be seen located near a5, a7, and a10 of the MyTH4 domain (Fig- showed a significant peak shift, consistent with the binding ure 4D), and these densities were shown to originate from showed by the ITC-based assays (Figure 3B). Moreover, the USH1G CEN2 (Wu et al., 2011). Interestingly, discontinuous smaller than the predicted elution volume for ANKS4B CEN in electron densities can also be seen at nearly the same positions the FPLC elution profile indicated that, similar to USH1G CEN of the MYO7B MyTH4 domain in the ANKS4B CEN/MYO7B (Wu et al., 2011), ANKS4B CEN also adopts an unfolded struc- NMFS complex (Figure 4C). Based on the high sequence similar- ture in solution (Figure 3B). We further separated CEN into two ities, we believe that these electron densities also originate from fragments, each containing the boxed CEN1 and CEN2 se- ANKS4B CEN2. quences shown in Figure 3C. ITC-based assay showed that To determine whether the ANKS4B/MYO7B and USH1G/ both fragments can bind to MYO7B NMFS but with weaker affin- MYO7A interactions are specific, we tested the bindings be- ity than the CEN region (Figure S5). Thus, both CEN1 and CEN2 tween MYO7A NMFS and ANKS4B CEN and between directly interact with MYO7B NMFS, a mode similar to what was MYO7B NMFS and USH1G CEN using ITC-based assays. No observed in the interaction between USH1G CEN and MYO7A binding could be detected in either of these combinations (Fig- NMFS (Wu et al., 2011)(Figure 3F). ure 3F), indicating that the two highly homologous MYO7s can

184 Developmental Cell 36, 179–189, January 25, 2016 ª2016 Elsevier Inc. specifically differentiate their respective targets USH1G and Given the current knowledge about the canonical target bind- ANKS4B, which in their own rights are also highly similar to ing modes of PDZ domains, the Harmonin NCPDZ3 and each other. Detailed analysis of the amino acid sequences MYO7B-CMF interaction is highly unusual, as there is no PBM and the two NMFS/CEN complex structures provides us with at the C terminus of MYO7B. The interaction is also unique for some hints regarding the binding specificity. Comparing the MyTH4-FERM tandems, as MyTH4-FERM tandems are not ANKS4B and USH1G CEN1 sequences, we find two differ- known to bind to a PDZ domain. We analyzed the atomic struc- ences: the corresponding residue of Leu265ANKS4B in ANKS4B ture of Harmonin apo-PDZ3 deposited in the PDB (PDB: 1V6B), is Phe307USH1G in USH1G (Figure 3C), which is conducive for and found that the sequence we used contains an N-terminal making additional hydrophobic contacts to Phe1473MYO7A extension forming a short a helix (N-extension) and a flexible and Ile1490MYO7A in MYO7A; Ser276ANKS4B in ANKS4B is C-terminal extension (C-extension, Figure S5). Both the N-exten- Arg318USH1G in USH1G (Figure 3C), which forms hydrogen sion and C-extension are conserved in Harmonins from fish to bonding with Ser1310MYO7A and charge-charge interaction mammals (Figure S5). Deletion of the N-extension mildly with Asp1317MYO7A. These two substitutions may explain, at decreased the binding affinity by 4-fold, while deletion of the least in part, the significantly stronger binding between C-extension totally abolished the binding (Figure 5E). However, MYO7A and USH1G than between MYO7B and ANKS4B no bindings could be detected for either the N-extension or the

(50 nM versus 1 mM in their Kd values). Our current structure C-extension alone (Figure 5E), indicating that the PDZ3 core pro- cannot provide details of the ANKS4B CEN2/MYO7B MyTH4 vides the major binding energy for CMF while the two extensions interaction, as we cannot define the structure of CEN2 based (the C-terminal extension in particular) facilitate the binding. In an on the very scarce electron densities. Moreover, although the earlier glutathione S-transferase (GST) pull-down-based map- overall architectures of the FERM domains in the CEN-bound ping study (Crawley et al., 2014b), PDZ3 was reported not to forms of the two MYO7 NMFS are quite similar (Figures 4C interact with the MYO7B tail. However, the sequence that the au- and 4D), the relative positions of the three lobes vary to some thors used (amino acids 360–534) lacked the entire C-terminal extent (the F2 and F3 lobes of MYO7B rotate 7 anti-clock- extension, which explains why no binding was detected. Thus, wise and 10 clockwise, respectively, when their F1 lobes the PDZ3 together with its N-extension and C-extension form are superimposed). Such conformational difference of the an integral extended PDZ domain to interact with MYO7B FERM domain cloverleaf structures may also contribute to the CMF. The detailed molecular basis governing this unexpected target recognition specificity of the two motors’ NMFS tan- complex formation awaits the determination of the atomic struc- dems. More broadly, the sensitivity of the target binding to ture of the complex. the inter-lobe conformation of the three lobes of a FERM To test whether the interaction between MYO7B CMF and domain can have potential impact on FERM domain-mediated Harmonin NCPDZ3 is specific for the brush border MYO7B, we target binding diversity and specificity as well as possible allo- used a GST pull-down assay to characterize the interaction be- steric regulations of FERM/target bindings. tween MYO7A and Harmonin. We used GST-tagged NCPDZ3 to pull down either the GFP-tagged MYO7A or GFP-tagged Interaction between MYO7B C-Terminal MyTH4-FERM MYO7B expressed in heterologous cells. MYO7B CMF can be and Harmonin PDZ3 robustly pulled down by GST-NCPDZ3, whereas only a small In stereocilia, MYO7A NMFS interacts with USH1G CEN with an amount of MYO7A CMF can be detected in the same assay (Fig- affinity around 50 nM, 20 times stronger than that between ure 5F). Thus, despite the high similarity between MYO7A CMF MYO7B NMFS and ANKS4B CEN. We next asked whether there and MYO7B CMF, the interaction between Harmonin NCPDZ3 might be additional binding sites for the MYO7B tail to enhance and MYO7B is highly specific. the link between actin filament bound myosin and the highly sta- ble ANKS4B/Harmonin complex. In an earlier study, it was re- DISCUSSION ported that CDHR2CD directly interacts with the MYO7B tail (Crawley et al., 2014b). We used ITC to quantify the binding be- In this study, we performed systematic biochemical and struc- tween CDHR2CD and the entire tail of MYO7B, and found the tural characterizations of the protein interaction network residing binding to be extremely weak (Kd 140 mM). Such weak interac- at the cytoplasmic face of the brush border inter-microvilli tip- tion between CDHR2CD and the MYO7B tail is not likely to be link. Our study includes detailed characterizations of previously physiologically meaningful, as the affinity between CDHR2PBM reported protein-protein interactions between CDHR2 and Har- and Harmonin PDZ2CC is about 100 times stronger (Figure 1B). monin and between MYO7B and Harmonin (Crawley et al., It was reported that Harmonin can also directly bind to the 2014b). We have also discovered in this study that ANKS4B, a MYO7B tail with an affinity stronger than the CDHR2/MYO7B protein highly homologous to USH1G and abundantly expressed interaction (Crawley et al., 2014b). Again, we used an ITC-based in brush border microvilli, functions as a key scaffold linking assay to verify this interaction, and found that the full-length Har- MYO7B with the Harmonin/CDHR2 complex. We have summa- monin binds to the MYO7B C-terminal MyTH4-FERM tandem rized the protein interaction network responsible for assembling with a Kd of 1.1 mM(Figure 5B). We further mapped the the brush border inter-microvillar tip-link complex in Figure 6A. MYO7B CMF binding region of Harmonin to be an extended The brush border inter-microvillar tip-link complex assembly is PDZ3 containing flanking sequences in both its termini (referred strikingly similar to the inner ear stereocilia tip-link complex to as NCPDZ3, Figures 5A–5C). FPLC-based binding assay organizations (Figures 6A and 6B). They both utilize adhesive, further confirmed the interaction between Harmonin NCPDZ3 heteromeric cadherin family proteins to build filamentous inter- and MYO7B CMF (Figure 5D). microvillar/inter-stereocilia tip-links that can sustain mechanical

Developmental Cell 36, 179–189, January 25, 2016 ª2016 Elsevier Inc. 185 Figure 5. The Extended PDZ3 Domain of Harmonin Binds to the MYO7B C-Terminal MyTH4-FERM Tandem (A) Domain organizations of Harmonin and MYO7B. (B) ITC-based assay shows that MYO7B CMF binds to the full-length Harmonin with moderate affinity. (C) ITC result showing that Harmonin PDZ3 containing both N- and C-terminal extension sequences had affinity toward MYO7B CMF similar to that of the full-length Harmonin. (D) FPLC-based assay showing that the extended PDZ3 of Harmonin and MYO7B CMF can form a complex in vitro. (E) Summary of ITC-derived binding results of different fragments of Harmonin and MYO7B, showing that both N- and C-extensions of Harmonin PDZ3 are required for Harmonin to bind to MYO7B CMF. (F) GST pull-down assay shows that Harmonin NCPDZ3 specifically binds to MYO7B CMF, but not to MYO7A CMF. strains. Both systems use two scaffold proteins (Harmonin and unique interaction features within each of the two systems. In ANKS4B in brush border, Harmonin and USH1G in inner ear) stereocilia, CDH23 interacts with Harmonin via multivalent inter- as the hub to organize the respective multi-protein complex actions using its PBM and residues encoded by exon 68 (Pan within each microvilli/stereocilia. The multi-domain scaffold pro- et al., 2009; Wu et al., 2012). The cytoplasmic tail of CDHR2 tein Harmonin appears to be particularly important. Via binding uses its PBM to interact with an extended PDZ2 domain to the CD of cadherins, Harmonin provides an anchoring point (PDZ2CC) of Harmonin. Interestingly, despite the above mecha- for the inter-microvillar/inter-stereocilia cadherin complex. nistic differences, the overall binding affinities of the CDH23/Har- Through binding to MYO7, Harmonin serves as a bridge con- monin complex and the CDHR2/Harmonin complex are rather necting the membrane-spanning cadherin complex with the similar. It should also be noted that the binding pocket located microvilli/stereocilia actin bundles (Figures 6A and 6B, bottom in Harmonin NTD is unoccupied in the microvilli tip-link complex, panels). suggesting that there might be additional Harmonin binding Nonetheless, there are clear distinctions between these two components remaining to be identified. In addition, the tripartite systems. First, except for Harmonin, which is present in both complex Harmonin/USH1G/MYO7A in stereocilia is formed by complexes, the rest of the proteins forming the microvilli tip- two pairs of very strong binary interactions between Harmonin link complex and those forming the stereocilia tip-link complex and USH1G and between USH1G and MYO7A (Figure 6B). In are different, although each protein in one system has a close ho- microvilli, the binary interaction between ANKS4B and Harmonin molog in the other. We have demonstrated in multiple cases that is also very strong. However, the interaction between ANKS4B the interactions within each system are often very specific and and MYO7B is much weaker than that between USH1G and not interchangeable. For example, the ANKS4B/MYO7B and MYO7A. Unique to the microvilli system, MYO7B forms another USH1G/MYO7A interactions are specific within each pair, pair of interactions with the extended PDZ3 of Harmonin, also despite the overall structures of the two complexes and their with moderate affinity. The simultaneous bindings of Harmonin interaction modes being highly similar. There also exists some and ANKS4B to MYO7B should ensure the formation of a highly

186 Developmental Cell 36, 179–189, January 25, 2016 ª2016 Elsevier Inc. Figure 6. Summary and Comparison of the Tip-Link Protein-Protein Interaction Networks in Brush Border Microvilli and Inner Ear Stereocilia (A) The top panel summarizes the detailed protein interaction network governing the assembly of the inter-microvillar tip-link. Except for the extracellular cadherin repeat-mediated hetero-dimerization interaction between CDHR2 and CDHR5 identified recently by Tyska’s group (Crawley et al., 2014b), the rest of the in- teractions marked by two-way arrows have been characterized quantitatively in this study. The bottom panel is a cartoon summary of the protein interaction network in microvilli. (B) The protein interaction network governing the inner ear stereocilia tip-link complex (also known as the USH1 complex) assembly. The figure is adapted from an earlier review by Pan and Zhang (2012). The bottom panel is a cartoon summary of the protein interaction network in stereocilia adapted from an earlier review by Lu et al. (2014). stable Harmonin/ANKS4B/MYO7B complex in microvilli. Both systems (Sekerkova et al., 2006). Eps8, an actin capping protein the overall similarities and unique differences in organizing the known to play essential roles in stereocilia elongation (Manor microvilli and stereocilia tip-link complexes presumably are et al., 2011), also is highly expressed in microvilli and is involved matched with the morphologically similar and functionally in the growth of microvilli. The striking similarities of the microvilli distinct properties of the two actin-based protrusions in animal and stereocilia systems suggest that knowledge obtained from kingdoms. one system might be valuable for understanding the other. At By comparing the proteomics studies between the brush present, our understanding of stereocilia tip-link complex is border and inner ear hair bundles (McConnell et al., 2011; Shin much more advanced than that of the brush border microvilli et al., 2013), these two actin-based protrusion systems share tip-link complex, at least partly due to clearly tractable clinical many more common/homologous components in addition to phenotypes and human genetics of the Ush1 syndromes, as the interaction networks discussed above. For example, several well as powerful animal models that can model human diseases cytoskeleton-related proteins are present in both systems. (Friedman et al., 2011; Gillespie and Muller, 2009; Richardson Myosin 1c is known to be located in stereocilia and is important et al., 2011). We anticipate that the systematic mechanistic- for adaptation of mechanical transductions (Gillespie and Cyr, based dissection of the brush border microvilli tip-link complex 2004); myosin 1a (also known as brush border myosin I) has organization reported in this study can provide valuable insights been reported to be important in brush border microvilli organi- for future functional studies of brush border microvilli in intestine zation (Tyska et al., 2005). Espin, a well-known actin bundling and kidney. We hope that the protein interaction network cha- protein, is important for the actin filament morphology in both racterized in our study may also be helpful in understanding

Developmental Cell 36, 179–189, January 25, 2016 ª2016 Elsevier Inc. 187 possible gut and/or kidney diseases caused by mutations of the et al., 2010). The final model was validated by MolProbity (Chen et al., 2010). genes encoding the proteins investigated in this work. The final refinement statistics are summarized in Table S1. All structure figures were prepared by PyMOL (http://www.pymol.org).

EXPERIMENTAL PROCEDURES Isothermal Titration Calorimetry Assay ITC measurements were carried out on a VP-ITC Microcal calorimeter (Micro- Constructs and Protein Expression cal) at 25C. Titration buffer contained 50 mM Tris-HCl (pH 7.8), 1 mM DTT, The Uniprot code of each protein and boundaries of recombinant proteins 1 mM EDTA, and 100 mM NaCl (low-salt condition) or 500 mM NaCl (high- used in this study are summarized in Table 1. The genes encoding CDHR2CD, salt condition). Each titration point was performed by injecting a 10-ml aliquot CDHR5CD, ANKS4B, and MYO7B NMFS were amplified by PCR from Mus of a protein sample from a syringe into a protein sample in the cell at a time in- musculus intestine cDNA and cloned into a pET vector. Fragments encoding terval of 120 s to ensure that the titration peak returned to the baseline. The Harmonin a and USH1G were amplified by PCR from Homo sapiens USH1C titration data were analyzed by Origin7.0 (Microcal). and USH1G, respectively, and cloned into a pET vector. Myosin VIIa fragments were PCR amplified from the mouse USH1B gene and cloned into a pET vector or pEGFP vector. The sequence encoding MYO7B CMF was amplified by PCR ACCESSION NUMBERS from the MYO7B tail construct provided by M.J. Tyska (Crawley et al., 2014b). Recombinant proteins were expressed in BL21 (DE3) Escherichia coli cells. The atomic coordinates and structure factors of the Harmonin/ANKS4B and 2+ ANKS4B/MYO7B complexes have been deposited at the PDB under the The N-terminal His6-tagged proteins were purified using an Ni -nitrilotriacetic acid agarose column followed by another step of size-exclusion chromatog- accession codes PDB: 5F3X and PDB: 5F3Y, respectively. raphy (Superdex 200 column from GE Healthcare) in the final buffer of 50 mM Tris-HCl, 1 mM DTT, 1 mM EDTA (pH 7.8), and 100 mM NaCl (low- SUPPLEMENTAL INFORMATION salt condition) or 500 mM NaCl (high-salt condition). GST-fused proteins were purified by GSH-Sepharose affinity chromatography, followed by a Supplemental Information includes six figures and one table and can be found step of size-exclusion chromatography the same as the one described above. with this article online at http://dx.doi.org/10.1016/j.devcel.2015.12.020.

GST Pull-Down Assay AUTHOR CONTRIBUTIONS For GST pull-down assays, the GFP-tagged MYO7A CMF and MYO7B CMF were individually expressed in HEK293 cells. GST-tagged NCPDZ3 or GST J.L., Y.H., Q.L., and M.Z. designed experiments and analyzed data. J.L. and (2 nmol in 0.9 ml of assay buffer composed of 50 mM Tris-HCl [pH 7.5], Y.H. performed experiments. J.L., Y.H., Q.L., and M.Z. wrote the manuscript. 100 mM NaCl, 1 mM EDTA, and 1 mM DTT) was incubated with 0.1 ml of M.Z. coordinated the research. HEK293 cell lysate for 1 hr at 4C. Next, each mixture was incubated with  30 ml of GSH-Sepharose 4B slurry beads for 30 min, also at 4 C. After washing ACKNOWLEDGMENTS two times, the captured proteins were eluted by boiling, resolved by 10% SDS- PAGE, and detected by immunoblotting with anti-GFP antibody (Santa Cruz We thank the Shanghai Synchrotron Radiation Facility (SSRF) BL17U and Biotechnology, GFP Antibody [B-2], catalog #sc-9996). BL19U1 beamlines for X-ray beam time. We are grateful to Dr. Matthew J. Tyska (Vanderbilt University) for providing the MYO7B CMF construct. This FPLC Coupled with Static Light Scattering work was supported by grants from RGC of Hong Kong (663811, 663812, Protein samples (typically 100 ml at a concentration of 20 mM pre-equilibrated 664113, AoE/M09/12, and T13-607/12R) to M.Z. M.Z. is a Kerry Holdings Pro- with corresponding column buffer) was injected into an AKTA FPLC system fessor of Science and a Senior Fellow of IAS at HKUST. with a Superose 12 10/300 Gl column (GE Healthcare) using the column buffer of 50 mM Tris-HCl (pH 7.8), 1 mM DTT, 1 mM EDTA, and 100 mM NaCl Received: October 29, 2015 (low-salt condition) or 500 mM NaCl (high-salt condition). The chromatography Revised: December 8, 2015 system was coupled to a static light scattering detector (miniDawn, Wyatt) and Accepted: December 21, 2015 differential refractive index detector (Optilab, Wyatt). The elution profiles were Published: January 25, 2016 analyzed using ASTRA 6 software (Wyatt). REFERENCES Crystallography Crystals of the ANKS4B SAM-PBM/Harmonin NPDZ1 complex (in 50 mM Tris Adams, P.D., Afonine, P.V., Bunkoczi, G., Chen, V.B., Davis, I.W., Echols, N., [pH 7.8], 500 mM NaCl, 1 mM EDTA, and 1 mM DTT buffer) and ANKS4B CEN/ Headd, J.J., Hung, L.W., Kapral, G.J., Grosse-Kunstleve, R.W., et al. (2010). MYO7B NMFS complex (in 50 mM Tris [pH 7.8], 100 mM NaCl, 1 mM EDTA, PHENIX: a comprehensive Python-based system for macromolecular struc- and 1 mM DTT buffer) were obtained by hanging-drop vapor diffusion methods ture solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213–221.  at 16 C. The crystals of the ANKS4B SAM-PBM/Harmonin NPDZ1 complex Barr-Gillespie, P.G. (2015). Assembly of hair bundles, an amazing problem for were grown in buffer containing 3.0 M sodium chloride, 0.1 M bis-Tris cell biology. Mol. Biol. Cell 26, 2727–2732. (pH 5.5), and soaked in crystallization solution containing 5.0 M sodium chlo- Bitner-Glindzicz, M., Lindley, K.J., Rutland, P., Blaydon, D., Smith, V.V., Milla, ride for cryoprotection. The crystals of ANKS4B CEN/MYO7B NMFS complex P.J., Hussain, K., Furth-Lavi, J., Cosgrove, K.E., Shepherd, R.M., et al. (2000). were grown in buffer containing 25% (w/v) pentaerythritol ethoxylate (3/4 EO/ A recessive contiguous gene deletion causing infantile hyperinsulinism, enter- OH) and 0.1 M 2-(N-morpholino)ethanesulfonic acid (pH 6.5), and soaked opathy and deafness identifies the Usher type 1C gene. Nat. Genet. 26, 56–60. in crystallization solution containing 40% (w/v) pentaerythritol ethoxylate Boeda, B., El-Amraoui, A., Bahloul, A., Goodyear, R., Daviet, L., Blanchard, S., (3/4 EO/OH) for cryoprotection. Diffraction data were collected at the Shanghai Perfettini, I., Fath, K.R., Shorte, S., Reiners, J., et al. (2002). Myosin VIIa, har- Synchrotron Radiation Facility (BL17U or BL19U1) at 100 K. Data were pro- monin and cadherin 23, three Usher I gene products that cooperate to shape cessed and scaled using HKL2000 (Otwinowski and Minor, 1997). the sensory hair cell bundle. EMBO J. 21, 6689–6699. Structure of the Harmonin NPDZ1/ANKS4B SAM-PBM complex was solved by molecular replacement with the model of USH1G/Harmonin complex (PDB: Chen, Z.Y., Hasson, T., Zhang, D.S., Schwender, B.J., Derfler, B.H., 3K1R) using PHASER (Mccoy et al., 2007). Structure of the ANKS4B CEN/ Mooseker, M.S., and Corey, D.P. (2001). Myosin-VIIb, a novel unconventional 72 MYO7B NMFS complex was also solved by molecular replacement with myosin, is a constituent of microvilli in transporting epithelia. Genomics , models of multiple fragments of the USH1G/MYO7A complex (PDB: 3PVL) 285–296. as the search model. Further manual model building and refinement were Chen, V.B., Arendall, W.B., Headd, J.J., Keedy, D.A., Immormino, R.M., completed iteratively using Coot (Emsley et al., 2010) and PHENIX (Adams Kapral, G.J., Murray, L.W., Richardson, J.S., and Richardson, D.C. (2010).

188 Developmental Cell 36, 179–189, January 25, 2016 ª2016 Elsevier Inc. MolProbity: all-atom structure validation for macromolecular crystallography. Nambiar, R., McConnell, R.E., and Tyska, M.J. (2010). Myosin motor function: Acta Crystallogr. D Biol. Crystallogr. 66, 12–21. the ins and outs of actin-based membrane protrusions. Cell. Mol. Life Sci. 67, Crawley, S.W., Mooseker, M.S., and Tyska, M.J. (2014a). Shaping the intesti- 1239–1254. nal brush border. J. Cell Biol. 207, 441–451. Otwinowski, Z., and Minor, W. (1997). Processing of X-ray diffraction data Crawley, S.W., Shifrin, D.A., Grega-Larson, N.E., McConnell, R.E., Benesh, collected in oscillation mode. Methods Enzymol. 276, 307–326. A.E., Mao, S.L., Zheng, Y.X., Zheng, Q.Y., Nam, K.T., Millis, B.A., et al. Pan, L.F., and Zhang, M.J. (2012). Structures of usher syndrome 1 proteins and (2014b). Intestinal brush border assembly driven by protocadherin-based in- their complexes. Physiology (Bethesda) 27, 25–42. termicrovillar adhesion. Cell 157, 433–446. Pan, L.F., Yan, J., Wu, L., and Zhang, M.J. (2009). Assembling stable hair cell El-Amraoui, A., and Petit, C. (2005). Usher I syndrome: unravelling the mech- tip link complex via multidentate interactions between harmonin and cadherin anisms that underlie the cohesion of the growing hair bundle in inner ear sen- 23. Proc. Natl. Acad. Sci. USA 106, 5575–5580. sory cells. J. Cell Sci. 118, 4593–4603. Richardson, G.P., de Monvel, J.B., and Petit, C. (2011). How the genetics of Emsley, P., Lohkamp, B., Scott, W.G., and Cowtan, K. (2010). Features and deafness illuminates auditory physiology. Annu. Rev. Physiol. 73, 311–334. development of Coot. Acta Crystallogr. D Biol. Crystallogr. 66, 486–501. Ridley, A.J. (2011). Life at the leading edge. Cell 145, 1012–1022. Friedman, T.B., Schultz, J.M., Ahmed, Z.M., Tsilou, E.T., and Brewer, C.C. (2011). Usher syndrome: hearing loss with vision loss. Adv. Otorhinolaryngol. Schwander, M., Kachar, B., and Muller, U. (2010). The cell biology of hearing. 70, 56–65. J. Cell Biol. 190, 9–20. Gillespie, P.G., and Cyr, J.L. (2004). Myosin-1c, the hair cell’s adaptation mo- Sekerkova, G., Zheng, L., Loomis, P.A., Mugnaini, E., and Bartles, J.R. (2006). tor. Annu. Rev. Physiol. 66, 521–545. Espins and the actin cytoskeleton of hair cell stereocilia and sensory cell 63 Gillespie, P.G., and Muller, U. (2009). Mechanotransduction by hair cells: microvilli. Cell. Mol. Life Sci. , 2329–2341. models, molecules, and mechanisms. Cell 139, 33–44. Shin, J.B., Krey, J.F., Hassan, A., Metlagel, Z., Tauscher, A.N., Pagana, J.M., Grati, M., and Kachar, B. (2011). Myosin VIIa and sans localization at stereocilia Sherman, N.E., Jeffery, E.D., Spinelli, K.J., Zhao, H.Y., et al. (2013). Molecular upper tip-link density implicates these Usher syndrome proteins in mechano- architecture of the chick vestibular hair bundle. Nat. Neurosci. 16, 365–374. transduction. Proc. Natl. Acad. Sci. USA 108, 11476–11481. Siemens, J., Kazmierczak, P., Reynolds, A., Sticker, M., Littlewood-Evans, A., Hirano, Y., Hatano, T., Takahashi, A., Toriyama, M., Inagaki, N., and and Muller, U. (2002). The Usher syndrome proteins cadherin 23 and harmonin Hakoshima, T. (2011). Structural basis of cargo recognition by the myosin-X form a complex by means of PDZ-domain interactions. Proc. Natl. Acad. Sci. MyTH4-FERM domain. EMBO J. 30, 2734–2747. USA 99, 14946–14951. Johnston, A.M., Naselli, G., Niwa, H., Brodnicki, T., Harrison, L.C., and Gonez, Sotomayor, M., Weihofen, W.A., Gaudet, R., and Corey, D.P. (2012). Structure L.J. (2004). Harp (harmonin-interacting, ankyrin repeat-containing protein), of a force-conveying cadherin bond essential for inner-ear mechanotransduc- a novel protein that interacts with harmonin in epithelial tissues. Genes Cells tion. Nature 492, 128–132. 9 , 967–982. Tilney, L.G., Tilney, M.S., and Derosier, D.J. (1992). Actin-filaments, stereocilia, Kazmierczak, P., Sakaguchi, H., Tokita, J., Wilson-Kubalek, E.M., Milligan, and hair-cells—how cells count and measure. Annu. Rev. Cell Biol. 8, 257–274. R.A., Muller, U., and Kachar, B. (2007). Cadherin 23 and protocadherin 15 Tyska, M.J., Mackey, A.T., Huang, J.D., Copeland, N.G., Jenkins, N.A., and interact to form tip-link filaments in sensory hair cells. Nature 449, 87–91. Mooseker, M.S. (2005). Myosin-1a is critical for normal brush border structure Kobayashi, I., Imamura, K., Yamada, M., Okano, M., Yara, A., Ikema, S., and and composition. Mol. Biol. Cell 16, 2443–2457. Ishikawa, N. (1998). A 75-kD autoantigen recognized by sera from patients Uhlen, M., Fagerberg, L., Hallstrom, B.M., Lindskog, C., Oksvold, P., with X-linked autoimmune enteropathy associated with nephropathy. Clin. Mardinoglu, A., Sivertsson, A., Kampf, C., Sjostedt, E., Asplund, A., et al. Exp. Immunol. 111, 527–531. (2015). Tissue-based map of the human proteome. Science 347, 1260419. Kobayashi, I., Imamura, K., Kubota, M., Ishikawa, S., Yamada, M., Tonoki, H., Verpy, E., Leibovici, M., Zwaenepoel, I., Liu, X.Z., Gal, A., Salem, N., Mansour, Okano, M., Storch, W.B., Moriuchi, T., Sakiyama, Y., et al. (1999). Identification A., Blanchard, S., Kobayashi, I., Keats, B.J.B., et al. (2000). A defect in harmo- of an autoimmune enteropathy-related 75-kilodalton antigen. Gastroenterology nin, a PDZ domain-containing protein expressed in the inner ear sensory hair 117,823–830. cells, underlies Usher syndrome type 1C. Nat. Genet. 26, 51–55. Louvard, D., Kedinger, M., and Hauri, H.P. (1992). The differentiating intestinal epithelial-cell—establishment and maintenance of functions through interac- Wang, C.K., Pan, L.F., Chen, J., and Zhang, M.J. (2010). Extensions of PDZ do- 1 tions between cellular structures. Annu. Rev. Cell Biol. 8, 157–195. mains as important structural and functional elements. Protein Cell , 737–751. Lu, Q., Li, J., and Zhang, M.J. (2014). Cargo recognition and cargo-mediated Wei, Z., Yan, J., Lu, Q., Pan, L., and Zhang, M.J. (2011). Cargo recognition regulation of unconventional myosins. Acc. Chem. Res. 47, 3061–3070. mechanism of myosin X revealed by the structure of its tail MyTH4-FERM tan- dem in complex with the DCC P3 domain. Proc. Natl. Acad. Sci. USA 108, Manor, U., Disanza, A., Grati, M., Andrade, L., Lin, H., Di Fiore, P.P., Scita, G., 3572–3577. and Kachar, B. (2011). Regulation of stereocilia length by myosin XVa and whir- lin depends on the actin-regulatory protein Eps8. Curr. Biol. 21, 167–172. Wu, L., Pan, L.F., Wei, Z.Y., and Zhang, M.J. (2011). Structure of MyTH4-FERM 331 Mathur, P., and Yang, J. (2015). Usher syndrome: hearing loss, retinal degen- domains in myosin VIIa tail bound to cargo. Science , 757–760. eration and associated abnormalities. Biochim. Biophys. Acta 1852, 406–420. Wu, L., Pan, L.F., Zhang, C.C., and Zhang, M.J. (2012). Large protein assem- McConnell, R.E., Benesh, A.E., Mao, S., Tabb, D.L., and Tyska, M.J. (2011). blies formed by multivalent interactions between Cadherin23 and harmonin Proteomic analysis of the enterocyte brush border. Am. J. Physiol. suggest a stable anchorage structure at the tip link of stereocilia. J. Biol. Gastrointest. Liver Physiol. 300, G914–G926. Chem. 287, 33460–33471. Mccoy, A.J., Grosse-Kunstleve, R.W., Adams, P.D., Winn, M.D., Storoni, L.C., Yan, J., Pan, L.F., Chen, X.Y., Wu, L., and Zhang, M.J. (2010). The structure of and Read, R.J. (2007). Phaser crystallographic software. J. Appl. Crystallogr. the harmonin/sans complex reveals an unexpected interaction mode of the 40, 658–674. two Usher syndrome proteins. Proc. Natl. Acad. Sci. USA 107, 4040–4045. Mooseker, M.S. (1985). Organization, chemistry, and assembly of the cyto- Ye, F., and Zhang, M.J. (2013). Structures and target recognition modes of skeletal apparatus of the intestinal brush-border. Annu. Rev. Cell Biol. 1, PDZ domains: recurring themes and emerging pictures. Biochem. J. 455, 209–241. 1–14.

Developmental Cell 36, 179–189, January 25, 2016 ª2016 Elsevier Inc. 189 Developmental Cell, Volume 36

Supplemental Information

Mechanistic Basis of Organization of the Harmonin/USH1C-Mediated Brush Border Microvilli Tip-Link Complex

Jianchao Li, Yunyun He, Qing Lu, and Mingjie Zhang

Figure S1. Biochemical characterization of the CDHR2/Harmonin and CDHR5/Harmonin interaction, related to Figure 1.

(A) ITC result showing that CDHR2CD binds to Harmonin with Kd ~5.65 µM. (B) ITC result showing that no interaction can be detected between CDHR5CD and Harmonin.

Figure S2. Comparison of the Harmonin NPDZ1/ANKS4B SAM-PBM complex structure and Harmonin NPDZ1/USH1G SAM-PBM, related to Figure 2. The two structures can be overlapped very well with RMSD of 0.71 Å.

Figure S3. ITC results between CDHR2 and Harmonin under high salt conditions or in the presence of ANKS4B SAM-PBM, related to Figure 2. (A) CDHR2CD can also bind to NPDZ1 with moderate affinity under high salt conditions. (B) Similar binding affinity can be detected between CDHR2CD and full length Harmonin under high salt condition. (C) The binding affinity decreased about 10 times between CDHR2 and PDZ2CC under high salt condition (1.4 µM vs 20.8 µM). (D) ANKS4B SAM-PBM prevent the binding between CDHR2CD and NPDZ1 under high salt condition. (E) ANKS4B SAM-PBM did not affect the binding between CDHR2CD and full length Harmonin.

Figure S4. Sequence alignment of NMFS of class VII myosin from different species, related to Figures 3&4. Identical and highly similar residues are colored (yellow for hydrophobic residues, blue for positively charged residues, red for negatively charged residues and green for other polar residues). Residues involved in CEN2 binding are highlighted with dash box. Residues involved in CEN1 binding are highlighted with black box.

Figure S5. ITC results between different fragments of ANKS4B CEN and MYO7B NMFS, related to Figure 3. ITC results showing that either CEN1 or CEN2 greatly decreased the binding affinity towards MYO7B NMFS.

Figure S6. Harmonin PDZ3 contains both N- and C-extension, related to Figure 5. (A) NMR structure of Harmonin PDZ3 showing that the N-extension forms a short α helix and the C-extension is highly flexible. (B) Sequence alignment showing that the PDZ3 together both the N- and C-extensions are highly conserved across different vertebrate species.

Table S1 Statistics of X-ray Crystallographic Data Collection and Model refinement, related to Figures 2&4.

Data collections Data sets Harmonin NPDZ1 MYO7B NMFS + ANKS4B SAM-PBM + ANKS4B CEN

Space group P 32 P 41212 Wavelength (Å) 0.9792 0.9788 Unit Cell Parameters (Å) a=b=124.57, c=49.65, a=b=197.52, c=97.71, α=β=90°, γ=120° α=β=γ=90° Resolution range (Å) 31.5-2.65 (2.70-2.65) 30.72-3.40 (3.46-3.40) No. of unique reflections 24978 (1264) 24972 (1262) Redundancy 5.7 (5.5) 6.8 (4.9) I/σ 38.2 (2.9) 19.6 (2.0) Completeness (%) 99.5 (100) 92.9 (95.7) a Rmerge (%) 6.2 (83.6) 10.0 (88.8)

Structure refinement Resolution (Å) 31.5-2.65 (2.77-2.65) 30.72-3.40 (3.55-3.40) b c Rcryst /Rfree (%) 21.72/26.60 (41.26/50.26) 24.13/28.05 (36.09/38.66) rmsd bonds (Å) / angles (°) 0.009/1.307 0.008/1.112 Average B factor (Å2) d 80.6 93.2 No. of atoms Protein atoms 4055 4041 Other molecules 21 0 No. of reflections Working set 23789 (3019) 23493 (2515) Test set 1133 (154) 1288 (132) Ramachandran plot regions d Favored (%) 97.5 96.7 Allowed (%) 2.5 3.1 Outliers (%) 0 0.2

Numbers in parentheses represent the value for the highest resolution shell. a. Rmerge = Σ |Ii - | / ΣIi, where Ii is the intensity of measured reflection and is the mean intensity of all symmetry-related reflections. b. Rcryst=Σ||Fcalc| – |Fobs||/ΣFobs, where Fobs and Fcalc are observed and calculated structure factors. c. Rfree= ΣT||Fcalc| – |Fobs||/ΣFobs, where T is a test data set of about 5% of the total unique reflections randomly chosen and set aside prior to refinement. d. B factors and Ramachandran plot statistics are calculated using MOLPROBITY (Chen et al., 2010).

References Chen, V.B., Arendall, W.B., Headd, J.J., Keedy, D.A., Immormino, R.M., Kapral, G.J., Murray, L.W., Richardson, J.S., and Richardson, D.C. (2010). MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr. D Biol. Crystallogr. 66, 12-21.